In a recent conversation, we were talking about the Kirkpatrick model, and a colleague had an interesting perspective that hadn’t really struck me overtly. Kirkpatrick is widely (not widely enough, and wrongly) used as an evaluation tool, but he talked about using it as a design tool, and that perspective made clear for me a problem with our approaches.
So, there’s a lot of debate about the Kirkpatrick model, whether it helps or hinders the movement towards good learning. I think it’s misrepresented (including by its own progenitors, though they’re working on that ;), and while I’m open to new tools I think it does a nice job of framing a fairly simple but important idea. The goal is to start with the end in mind.
And the evidence is that it’s not being used well. The largest implementation of the model is level 1, which isn’t of use (correlation between learner reaction and actual impact is .09, essentially zero with a rounding error). Level 2 drops to a third of orgs, and it drops from there. And this is broken.
The point, and this is emphasized by the ‘design’ perspective, is that you are supposed to start with level 4, and work back. What’s the measurable indicator in the organization that isn’t up to snuff, and what behavior (level 3) would likely impact that? And how do we change that behavior (Level 2)? And here’s where it can go beyond training: that intervention might be a job aid, or access to a network (which hasn’t been much in the promotion of the model).
To be fair, the proponents do argue you should be starting at Level 4, but with the numbering (which Don admits he might have got wrong) and the emphasis on evaluation, it doesn’t hit you up front. Using it as a design tool, however, would emphasize the point.
So here’s to thinking of learning design as working backwards from a problem, not forwards from a request. And, of course, to better learning design overall.